public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Joerg Roedel <joerg.roedel@amd.com>
To: Ingo Molnar <mingo@elte.hu>
Cc: iommu@lists.linux-foundation.org, linux-kernel@vger.kernel.org,
	Joerg Roedel <joerg.roedel@amd.com>
Subject: [PATCH 28/34] x86/amd-iommu: Add device bind reference counting
Date: Fri, 27 Nov 2009 14:55:39 +0100	[thread overview]
Message-ID: <1259330145-14865-29-git-send-email-joerg.roedel@amd.com> (raw)
In-Reply-To: <1259330145-14865-1-git-send-email-joerg.roedel@amd.com>

This patch adds a reference count to each device to count
how often the device was bound to that domain. This is
important for single devices that act as an alias for a
number of others. These devices must stay bound to their
domains until all devices that alias to it are unbound from
the same domain.

Signed-off-by: Joerg Roedel <joerg.roedel@amd.com>
---
 arch/x86/include/asm/amd_iommu_types.h |    1 +
 arch/x86/kernel/amd_iommu.c            |   37 +++++++++++++++++++++++++-------
 2 files changed, 30 insertions(+), 8 deletions(-)

diff --git a/arch/x86/include/asm/amd_iommu_types.h b/arch/x86/include/asm/amd_iommu_types.h
index 9eaa27b..434e90e 100644
--- a/arch/x86/include/asm/amd_iommu_types.h
+++ b/arch/x86/include/asm/amd_iommu_types.h
@@ -253,6 +253,7 @@ struct protection_domain {
 struct iommu_dev_data {
 	struct device *alias;		  /* The Alias Device */
 	struct protection_domain *domain; /* Domain the device is bound to */
+	atomic_t bind;			  /* Domain attach reverent count */
 };
 
 /*
diff --git a/arch/x86/kernel/amd_iommu.c b/arch/x86/kernel/amd_iommu.c
index 3214e88..f5db7d5 100644
--- a/arch/x86/kernel/amd_iommu.c
+++ b/arch/x86/kernel/amd_iommu.c
@@ -152,6 +152,8 @@ static int iommu_init_device(struct device *dev)
 	if (pdev)
 		dev_data->alias = &pdev->dev;
 
+	atomic_set(&dev_data->bind, 0);
+
 	dev->archdata.iommu = dev_data;
 
 
@@ -1403,10 +1405,13 @@ static int __attach_device(struct device *dev,
 		return -EBUSY;
 
 	/* Do real assignment */
-	if (alias != devid &&
-	    alias_data->domain == NULL) {
-		alias_data->domain = domain;
-		set_dte_entry(alias, domain);
+	if (alias != devid) {
+		if (alias_data->domain == NULL) {
+			alias_data->domain = domain;
+			set_dte_entry(alias, domain);
+		}
+
+		atomic_inc(&alias_data->bind);
 	}
 
 	if (dev_data->domain == NULL) {
@@ -1414,6 +1419,8 @@ static int __attach_device(struct device *dev,
 		set_dte_entry(devid, domain);
 	}
 
+	atomic_inc(&dev_data->bind);
+
 	/* ready */
 	spin_unlock(&domain->lock);
 
@@ -1449,20 +1456,34 @@ static int attach_device(struct device *dev,
  */
 static void __detach_device(struct device *dev)
 {
-	u16 devid = get_device_id(dev);
+	u16 devid = get_device_id(dev), alias;
 	struct amd_iommu *iommu = amd_iommu_rlookup_table[devid];
 	struct iommu_dev_data *dev_data = get_dev_data(dev);
+	struct iommu_dev_data *alias_data;
 
 	BUG_ON(!iommu);
 
-	clear_dte_entry(devid);
-	dev_data->domain = NULL;
+	devid = get_device_id(dev);
+	alias = get_device_id(dev_data->alias);
+
+	if (devid != alias) {
+		alias_data = get_dev_data(dev_data->alias);
+		if (atomic_dec_and_test(&alias_data->bind)) {
+			clear_dte_entry(alias);
+			alias_data->domain = NULL;
+		}
+	}
+
+	if (atomic_dec_and_test(&dev_data->bind)) {
+		clear_dte_entry(devid);
+		dev_data->domain = NULL;
+	}
 
 	/*
 	 * If we run in passthrough mode the device must be assigned to the
 	 * passthrough domain if it is detached from any other domain
 	 */
-	if (iommu_pass_through)
+	if (iommu_pass_through && dev_data->domain == NULL)
 		__attach_device(dev, pt_domain);
 }
 
-- 
1.6.5.3



  parent reply	other threads:[~2009-11-27 14:00 UTC|newest]

Thread overview: 36+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2009-11-27 13:55 [git pull] IOMMU updates for 2.6.33 Joerg Roedel
2009-11-27 13:55 ` [PATCH 01/34] x86/amd-iommu: Separate internal interface definitions Joerg Roedel
2009-11-27 13:55 ` [PATCH 02/34] x86/amd-iommu: Update copyright headers Joerg Roedel
2009-11-27 13:55 ` [PATCH 03/34] x86/amd-iommu: Add an index field to struct amd_iommu Joerg Roedel
2009-11-27 13:55 ` [PATCH 04/34] x86/amd-iommu: Add per IOMMU reference counting Joerg Roedel
2009-11-27 13:55 ` [PATCH 05/34] x86/amd-iommu: Add function to complete a tlb flush Joerg Roedel
2009-11-27 13:55 ` [PATCH 06/34] x86/amd-iommu: Make iommu_flush_pages aware of multiple IOMMUs Joerg Roedel
2009-11-27 13:55 ` [PATCH 07/34] x86/amd-iommu: Use __iommu_flush_pages for tlb flushes Joerg Roedel
2009-11-27 13:55 ` [PATCH 08/34] x86/amd-iommu: Remove iommu_flush_domain function Joerg Roedel
2009-11-27 13:55 ` [PATCH 09/34] x86/amd-iommu: Implement protection domain list Joerg Roedel
2009-11-27 13:55 ` [PATCH 10/34] x86/amd-iommu: Reimplement amd_iommu_flush_all_domains() Joerg Roedel
2009-11-27 13:55 ` [PATCH 11/34] x86/amd-iommu: Reimplement flush_all_domains_on_iommu() Joerg Roedel
2009-11-27 13:55 ` [PATCH 12/34] x86/amd-iommu: Make np-cache a global flag Joerg Roedel
2009-11-27 13:55 ` [PATCH 13/34] x86/amd-iommu: Use check_device for amd_iommu_dma_supported Joerg Roedel
2009-11-27 13:55 ` [PATCH 14/34] x86/amd-iommu: Use check_device in get_device_resources Joerg Roedel
2009-11-27 13:55 ` [PATCH 15/34] x86/amd-iommu: Remove iommu parameter from dma_ops_domain_(un)map Joerg Roedel
2009-11-27 13:55 ` [PATCH 16/34] x86/amd-iommu: Make alloc_new_range aware of multiple IOMMUs Joerg Roedel
2009-11-27 13:55 ` [PATCH 17/34] x86/amd-iommu: Remove iommu parameter from __(un)map_single Joerg Roedel
2009-11-27 13:55 ` [PATCH 18/34] x86/amd-iommu: Remove iommu specific handling from dma_ops path Joerg Roedel
2009-11-27 13:55 ` [PATCH 19/34] x86/amd-iommu: Let domain_for_device handle aliases Joerg Roedel
2009-11-27 13:55 ` [PATCH 20/34] x86/amd-iommu: Simplify get_device_resources() Joerg Roedel
2009-11-27 13:55 ` [PATCH 21/34] x86/amd-iommu: Move find_protection_domain to helper functions Joerg Roedel
2009-11-27 13:55 ` [PATCH 22/34] x86/amd-iommu: Use get_device_id and check_device where appropriate Joerg Roedel
2009-11-27 13:55 ` [PATCH 23/34] x86/amd-iommu: Remove iommu parameter from dma_ops_domain_alloc Joerg Roedel
2009-11-27 13:55 ` [PATCH 24/34] x86/amd-iommu: Move some pte allocation functions in the right section Joerg Roedel
2009-11-27 13:55 ` [PATCH 25/34] x86/amd-iommu: Rearrange dma_ops related functions Joerg Roedel
2009-11-27 13:55 ` [PATCH 26/34] x86/amd-iommu: Remove support for domain sharing Joerg Roedel
2009-11-27 13:55 ` [PATCH 27/34] x86/amd-iommu: Use dev->arch->iommu to store iommu related information Joerg Roedel
2009-11-27 13:55 ` Joerg Roedel [this message]
2009-11-27 13:55 ` [PATCH 29/34] x86/amd-iommu: Keep devices per domain in a list Joerg Roedel
2009-11-27 13:55 ` [PATCH 30/34] x86/amd-iommu: Cleanup attach/detach_device code Joerg Roedel
2009-11-27 13:55 ` [PATCH 31/34] x86/amd-iommu: Introduce iommu_flush_device() function Joerg Roedel
2009-11-27 13:55 ` [PATCH 32/34] x86/amd-iommu: Cleanup DTE flushing code Joerg Roedel
2009-11-27 13:55 ` [PATCH 33/34] x86/amd-iommu: Move reset_iommu_command_buffer out of locked code Joerg Roedel
2009-11-27 13:55 ` [PATCH 34/34] x86/amd-iommu: Remove amd_iommu_pd_table Joerg Roedel
2009-11-27 19:27 ` [git pull] IOMMU updates for 2.6.33 Ingo Molnar

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1259330145-14865-29-git-send-email-joerg.roedel@amd.com \
    --to=joerg.roedel@amd.com \
    --cc=iommu@lists.linux-foundation.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@elte.hu \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox